The corporate innovation field guide

Who this is for

Innovation leads. CX leads. Digital and product leads. HR and L&D leads who just inherited "capability." CIO, CDIO, CDO, and CAIO-shaped owners who still have to explain a quarter.

You sit at ANZ, NAB, Telstra, AGL, DBS, Allens, an NHS trust, Services Australia, a NSW cluster, IRD in New Zealand, a GDS-shaped UK team, Capital One, Atlassian, Canva, or a Canadian operator like Inkeros. The letterhead changes. The physics do not.

You have already sat through the offsite. You can buy Design Thinking, Customer Experience, Lean Startup and Product Sprints, Productivity and Time Management, AI Training, Agile, or AI Design Sprints. You do not need another mural. You need residue: a framed problem, a funded test, a named owner who was in the room, and a kill date a risk lead can sign.

If you cannot name those four before you book the room, do not book the room. This guide is the anti-playbook for that sentence.

How to use this (it is a field guide)

Print the canvases. Fill them in ink. Circulate the residue the same day. Do not turn this into a 40-page pack.

Work it in this order:

1. Retire the theatre (anti-playbook).
2. Fill the 30-day experiment canvas, including kill criteria, before you send a calendar invite.
3. Brief the workshop so it ships a decision, not a photo.
4. Steal a named-account pattern from a peer geo (AU, NZ, HK, SG, US, UK, Canada). Do not invent their numbers.
5. Buy the workshop that matches the job. The pillar map is later in this guide.
6. Run the Monday residue checklist before people reach the lift.

If a section does not change what you do on Tuesday, skip it. Residue is the only honest unit of progress.

The anti-playbook (theatre to retire)

Corporate innovation has a branding problem. The logo is a lightbulb. The artefact is a mural. The outcome is a slide. We spent a decade teaching large organisations in finance, government, legal, healthcare, energy, and product to adopt the language of startups without adopting the constraints that make startups honest. Startups die. Your lab does not. That is the whole difference.

Collective Campus runs workshops. We are against workshops that leave no residue. The don'ts below are the part most executive teams have not written down. The dos are already in the vendor deck. The don'ts protect a quarter.

Theatre to retire this quarter

• A lab with no killing mechanism. That is a hobby with better catering.
• Training 400 people while last year's KPI still pays the bonus.
• A hackathon whose follow-up budget is "we will find funding." If nobody owns Monday's money, you ran a party.
• An AI strategy bought off a vendor deck. Start with a calendar, not a capability map.
• A journey map no operator will hang above their desk.
• An offsite confused with a strategy. A strategy is a set of bets with owners, kill dates, and a resource shift. An offsite is a calendar invite.
• Risk, legal, procurement, and comms invited in week nine as a surprise. They are users. Interview them before you mural them.
• A centre of excellence that presents the demo while the desk was not in the room.
• Completions, logins, licences, and facilitator NPS reported as outcomes.
• A pilot that never expires. The NHS pattern that fails is the pilot with no stop. Copy the failure and you will fund a museum.

What residue looks like instead

One framed problem in one sentence. One constraint in one sentence. One thinnest test the constraint will allow (concierge, paper, shadow process, throwaway page). One named owner who sat in the room. One kill date. One metric (hours, cycle time, conversion, complaints). One page circulated the same day.

If every idea graduates, none of them were experiments.

A practical close: put the kill date on the same page as the hypothesis. If you cannot name the kill, you do not have a bet. You have a hope.

30-day experiment canvas

Fill this on one page. If it spills to two, you are writing a business case. Stop.

The canvas (copy onto a page)

1. Hated workflow or customer seam (one sentence). Not "improve onboarding." "The small-business pack that stalls for ten days because we re-key data we already have." Not "better CX." "The bill explanation a retail lead cannot defend on a hardship call."

2. Who hurts (role, not persona poster). Credit officer. Care writer. Partner's EA. Frontline claims officer. Store lead. Taxpayer intermediary. Name a desk.

3. Constraint that usually kills the idea (one sentence). APRA or ASIC. MAS or HKMA. AER retail rules. Clinical safety. Information Governance. Partner leverage. Industrial relations. Procurement threshold. Comms embargo. Write it on day one.

4. Thinnest test the constraint will allow. Concierge. Paper script. Shadow process. Manual decision with a clock. Throwaway landing page. Redesigned letter. Not a platform. Not a core change.

5. Hypothesis (if / then / because). If we run [test] with [cohort] for [days], then [observable] will move, because [reason we can be wrong].

6. Metric (one number). Hours returned. Cycle time. Completion. Complaint volume. Partner time on intake. Pick one. Secondary metrics go in a footnote, not in the headline.

7. Named owner (a person, not a centre). The person who can move a roster, a backlog, or a budget. They must attend. A delegate who cannot kill is applause.

8. Named stopper (risk, legal, clinical, credit, or ops). The person who will stop you. Give them the page before the room. They mark it. They sit in the last hour.

9. Dollar figure and what it buys. Room, time, test channel, concierge hours, page, ads, or letters. A number a CFO can see without a pack.

10. Kill date and kill criteria. A calendar day, not a vibe. Thirty days is a default, not a religion. Write the stop in language a risk lead and a CFO can both sign.

11. What we will stop doing if this lives. If you cannot name a meeting, a report, a step, or a template that dies, you bought capacity for more theatre.

12. Monday residue owner. Who sends the one-pager before people reach the lift. Same day. Not "comms will tidy it Friday."

Worked canvas shapes (patterns, not invented scores)

ANZ or NAB (AU / NZ banking). Seam: a decision explanation or a small-business document chase. Constraint: operational risk and conduct. Test: shadow process or a rewritten explanation a credit and comms lead will sign. Owner: product plus ops. Stopper: operational risk. Metric: cycle time or rework hours. Kill: if the shadow cannot complete a capped cohort without a control exception.

Telstra (AU telco). Seam: care or field explanation rebuilt from a blank page. Constraint: customer data and brand. Test: a play plus a checked first draft, not a new knowledge platform. Owner: care or field lead. Stopper: privacy and comms. Metric: minutes per explanation, then hours at day 14 and 30. Kill: if the check is skipped or the minutes do not move.

AGL (AU energy). Seam: a retail offer or a hardship conversation. Constraint: AER-shaped retail rules and a household bill you cannot "move fast and break." Test: offer script or concierge, not a new product in the stack. Owner: retail. Stopper: retail risk. Metric: comprehension or completion of the scripted path. Kill: if claims or hardship treatment fail a risk mark.

DBS (SG bank). Seam: a desk pack or a credit memo. Constraint: MAS-shaped controls. Test: a play a domain lead and a builder will run on Tuesday. Owner: the desk. Stopper: operational risk. Metric: hours returned. Kill: if the play cannot sit next to a control.

Allens (AU legal, Linklaters alliance). Seam: intake, scoping, or knowledge handoff. Constraint: partner time and matter risk. Test: a first-client email or a scoping note a partner will fund. Owner: a partner plus an operator. Stopper: general counsel-shaped risk on the matter type. Metric: partner minutes on intake. Kill: if the practice will not staff the new path on a live matter.

NHS trust or GDS-shaped UK team. Seam: a pathway, a letter, or a transaction a citizen cannot finish. Constraint: clinical safety, IG, or the service standard. Test: paper script or a service slice, not a spine system. Owner: service owner. Stopper: IG or clinical safety. Metric: completion or time to a safe next step. Kill: if the test needs an act rewritten by Christmas.

Services Australia or NSW cluster (AU public). Seam: identity, payments, or a "make it easier" handoff that dies between clusters. Constraint: privacy, legislation, volume. Test: concierge or a permissioned 90-day slice. Owner: operational owner. Stopper: privacy and security. Metric: completion of the slice. Kill: if the constraint arrives as a surprise in week nine, you framed it wrong. Reframe or stop.

IRD (NZ). Seam: a letter or an intermediary journey. Constraint: the act. Test: a rewritten letter a policy owner will sign. Owner: service plus policy. Metric: contact rate or completion. Kill: if the letter cannot survive the act.

Capital One (US). Seam: a product or servicing explanation. Constraint: credit and conduct. Test: design-org spike with evidence as a reason to stop, not only a reason to scale. Owner: product. Stopper: credit or conduct. Metric: a customer behaviour, not a mural photo.

Atlassian or Canva (product companies, AU / global). Seam: an internal workflow that still pastes and chases, or a customer path the squad already owns. Constraint: customers who will reply in public, plus data. Test: a spike or a play inside an existing heartbeat. Owner: the squad. Metric: hours or cycle time. Kill: if it adds volume into the same review with no backlog discipline.

Inkeros (Canada / US). Seam: a sales or engagement workflow the room already hates. Constraint: omnichannel reality, not a lab. Test: two-day framing then a shortlist with owners. Owner: a line lead who can fund. Kill: if the shortlist has no names on Monday.

Infinitev (AU / NZ). Seam: an offer you would otherwise wait a quarter to test. Constraint: claims, channel, brand. Test: a throwaway landing page in days. Owner: commercial. Stopper: claims. Metric: pulse (enquiry, not vanity traffic). Kill: if the page cannot live inside the claims rule.

Different logos. Same canvas.

Kill criteria a risk lead can sign

Write the stop before the concept. Boards relax when they can see the kill. They tense when they can only see the dream.

Use this language. Edit the nouns. Do not soften the verb "stop."

• We will stop if we cannot get the capped cohort through the concierge or paper version by the kill date.
• We will stop if the unit economics require a subsidy the board will not name.
• We will stop if legal, IG, clinical safety, or credit will not let us run the paper version.
• We will stop if the named owner misses two heartbeats and no deputy was named.
• We will stop if the metric does not move enough to pay for the next 30 days (write the "enough," even if it is ugly).
• We will stop if the play requires a production integration on day one.
• We will stop if comms will not sign the public sentence.

ANZ-style and NAB-style programmes that survive do this in public, inside the team. They do not hide the kill in a stage-gate that only fires after a large cheque. Capital One's better design work treated evidence as a reason to stop. The AGL pattern that works is the offer test a retail and risk lead can live with, then kill without a press release. DBS-style digital work treated owners and P&L, not murals. GDS-style assessments already punish fiction. Steal the punishment.

Public stop. If you kill it, write the learning in the same channel you would have written the win. That is how you earn the next test.

How to brief a workshop so it ships a decision

A workshop that ships a decision was briefed like a product spike. A workshop that ships a mural was briefed like a town hall.

Do not send an invite titled "co-creation" or "ideation day." Send a brief. Twenty minutes to read. One page plus the canvas.

The brief (steal the headings)

Decision we need by close of day two. Go, iterate, or kill on [named test]. Not "align on opportunities."

Problem in four lines. Customer or desk. Constraint. Why now. What we already tried.

Who must be in the room (and who must not). Decision-maker who can kill. Named owner. Named stopper (risk, legal, procurement, comms, clinical, credit). Two people who do the work. One builder or BA. Facilitator. That is the room. Do not add a gallery.

Pre-work (90 minutes, not a reading pack). Owner fills draft canvas. Stopper marks the constraint. One operator writes the current steps with times. Bring three artefacts (ticket, letter, bill, memo, intake email). No persona homework.

Artefacts we will leave. Filled canvas. Test design. Risk note (one page). Comms sentence. Owner, stopper, kill date, dollar figure. Residue note sent the same day.

What we will not do in the room. Icebreakers that steal an hour. Framework tours. Roadmaps to 2029. Voting on twenty ideas. A "next steps" slide with no names.

Aftercare home. ANZ and NAB squads can hang this on an existing heartbeat. Telstra and AGL can hang it on a journey or retail forum. Allens can hang it on a practice meeting. Services Australia and NSW can hang it on a programme board. GDS-shaped teams can hang it on an assessment window. Inkeros can hang it on a sales leadership meeting. If aftercare has no home, you will buy a sequel. Sequels are how theatre bills you twice.

Invite list that actually ships

Invite the person who can say no. Invite the person whose Tuesday changes. Invite the person who will stop you. If any of the three sends a note-taker, reschedule.

Do not invite:

• A sponsor who will keynote and leave
• Three vendors
• A centre of excellence with no desk
• Comms "for awareness" if they will not mark a sentence
• Twelve teams (that is a hackathon)

Inkeros-style two-day sprints work when day one is framing and day two is prototypes plus decision criteria, and the shortlist has owners. Infinitev-style rooms work when claims sit in the chair before the page is drafted. Atlassian-style ShipIt patterns only pay when a path back to a product owner exists before the venue is booked. Copy those shapes. Do not copy a festival.

Facilitator instruction (one paragraph)

Your job is a decision and a page. Time-box disagreement. Put the constraint on the wall in the first hour. Prototype the paper trail with the same seriousness as the interface. Book the decision-maker for the last afternoon. If they cannot attend, you scheduled a lab. Circulate residue before people stand up.

Named-account patterns (steal the physics, not a fake quote)

We name companies so a buyer in the geo has a reason to visit, then a play they can run. We do not invent quotes. We do not invent their metrics. The org chart, the licence, and the public programme shape are enough.

Banking (ANZ, NAB, DBS, Capital One)

ANZ has spent years on agile and design-led work across Australia and New Zealand. The useful part is not the framework badge. The useful part is when a squad can kill a feature without a steering pack that restates the workshop.

NAB is an Authorised Deposit-taking Institution. Experiments fail when they start as a core-banking change. They succeed when they start as a shadow process or a manual decision with a clock. Operational risk can mark a one-page control. They cannot mark a mural.

DBS is routinely held up as the digital bank that left the pack. Fair. The useful part is that digital work was treated as a P&L problem with owners. The current chapter is making tools a desk conversation without turning every officer into shadow IT.

Capital One built a genuine design org in the US and still had to fight the gravitational pull of the slide. Evidence as a reason to stop is the steal.

HK and SG peers (OCBC, HSBC Hong Kong) rhyme on controls (MAS, HKMA). Same canvas. Tighter guardrail page if the workshop is AI.

Telco, energy, legal (Telstra, AGL, Allens)

Telstra has run digital and AI programmes that would fill a conference. The useful tell is whether care, field, and corporate teams get a sequence (guardrails, use cases, practice) or a portal and a hope.

AGL has to invent inside a listed energy business. The thing you break is a household bill or a generator. Lean inside AGL-shaped retail is an offer test, a script, or a concierge a retail and risk lead can live with.

Allens, through the Linklaters alliance, feels client seams in intake, scoping, and knowledge handoff. Partners will not fund a journey map. They will fund a change that fits a matter if a partner and an operator sat in the room. Legal is not allergic to design. Legal is allergic to design that cannot survive a file.

Public sector (NHS, GDS, Services Australia, NSW, IRD)

The NHS has more service-design talent per square metre than most consultancies, and still loses years to business cases that pretend certainty. Write a 90-day window and a kill date. Do not write a spine replacement as the first test.

GDS in the UK made the useful questions boring: who is the user, what is the need, what have you learned, what will you do next, what will you stop? Steal the questions even if you are not a department.

Services Australia lives inside legislation, identity, and volume. Useful work is problem framing that can survive privacy, security, and payments on the same page.

NSW has Service NSW as the public face of "make it easier" and cluster owners behind it. Pattern that works: operational owner, 90-day window, kill date. Pattern that fails: a lab mural and a business case that pretends the act will be rewritten by Christmas.

IRD NZ work that moves treats the taxpayer (or intermediary) and the legislation as the same brief.

Product and growth operators (Atlassian, Canva, Inkeros, Infinitev)

Atlassian already has sprint muscle and a ShipIt tradition. That only matters when the path to a backlog exists.

Canva already lives inside the tools it sells. The internal question is still which workflows get hours back this month.

Inkeros (Canada and the United States) is the two-day sprint pattern: frame, prototype, decision criteria, shortlist with owners.

Infinitev (Australia and New Zealand) is the offer-test pattern: AI-assisted landing pages so the team could test ideas in days, not quarters. Claims stay in the room.

Geo note

Opportunities outreach stays tight on AU, NZ, HK, SG. This newsroom also writes US, UK, and Canada. The canvas does not change when the flight does. The stopper's letterhead does.

When to buy which workshop (pillar map)

Collective Campus T&E is not a tasting menu. Buy the format that matches the job. If you buy the wrong pillar you will get a good room and a useless Monday.

Design Thinking

Buy when: the problem is framed badly, users (including risk and procurement) disagree about the job, and you need a tested concept plus a decision owner.

Do not buy when: you already know the workflow and you need hours back (that is AI Training or Productivity) or you need a cadence (that is Agile).

Residue: framed problem, concept, kill criteria, owner. See the companion primer.

Customer Experience

Buy when: operators will not use the current map, the seam crosses channels, and you need a journey a roster can run.

Do not buy when: you want a brand film or a persona poster for the wall.

Residue: a map with times, owners, and one broken step you will fix in 30 days.

Lean Startup and Product Sprints

Buy when: you need a test licence inside a regulated business (AGL-shaped energy, NAB-shaped banking, NHS-shaped care). Build, measure, learn, with a cap and an expiry.

Do not buy when: you want a platform business case.

Residue: test licence, cohort cap, kill date, learning you will publish internally if it dies.

Productivity and Time Management

Buy when: the calendar is the product. Meetings, maker time, and a hated weekly pack.

Do not buy when: you want a wellness poster.

Residue: deleted rooms, a working agreement, hours back on one pack.

AI Training

Buy when: a team already hates a workflow, you can pair a domain lead with a builder, and a risk lead will sign a guardrail page.

Do not buy when: you want a model zoo, a licence dump, or a town-hall video.

Residue: a play, hours at day 14 and 30, keep / kill / constrain. See the companion playbook.

Agile

Buy when: the team has no heartbeat, cannot say no, and demos to nobody who can kill.

Do not buy when: you want another certification drive or a scaling framework tour.

Residue: a two-week cadence, a backlog with a no, a sceptic in the demo.

AI Design Sprints

Buy when: you need a decision and a throwaway artefact next week (Infinitev-style page, script, letter), and a decision-maker can sit in the last afternoon.

Do not buy when: you want twelve teams and a trophy (that is a hackathon) or you want literacy on a desk play (that is AI Training).

Residue: go / iterate / kill plus an artefact you can put in front of a customer or a steering owner.

Quick chooser

• Bad problem, many users, need a concept: Design Thinking.
• Broken service path operators ignore: Customer Experience.
• Regulated test with a cap: Lean Startup.
• Calendar is on fire: Productivity.
• Hated workflow, hours, guardrails: AI Training.
• No cadence, cannot say no: Agile.
• Need a test artefact this week: AI Design Sprint.
• Saturday energy, no owner: do not buy. Fix ownership on a Wednesday.

If you are a CIO funding two of these in the same quarter, sequence them. Design Thinking or Lean frames the bet. Agile gives it a heartbeat. AI Training or Productivity takes hours back on the desk that has to live with the bet. Do not run all seven as a festival.

Monday residue checklist

Run this before people reach the lift. If you cannot tick it, you ran stationery.

• Problem in one sentence, in the words of the desk or the customer.
• Constraint in one sentence, in the words of the stopper.
• Thinnest test named (concierge, paper, shadow, page, letter).
• Named owner who was in the room (email, not a function).
• Named stopper who marked the page.
• Kill date on a calendar.
• Kill criteria a risk lead and a CFO can both sign.
• Dollar figure for the next 30 days.
• One metric, with a baseline even if the baseline is ugly.
• What we will stop doing if this lives.
• Aftercare home (heartbeat, forum, practice meeting, programme board).
• One-page residue note sent the same day. Not a pack. Not "we will write up next week."
• Decision-maker sentence they will repeat in the board pack.
• Next room booked only if this test needs it. No sequel for its own sake.

Print this. Tape it to the facilitator wall. Read it out loud at 4pm on day two.

Don'ts (because someone will try them anyway)

Don't launch a lab with no killing mechanism. Write the stop first.

Don't train a cohort and change zero incentives. Pick one team. Change one metric for one quarter. Pair the T&E with that metric.

Don't run a hackathon to "build culture." Culture is a wet Tuesday. Book the follow-up before the venue.

Don't buy an AI strategy off a vendor deck. Start with a calendar.

Don't separate lawyers, risk leads, procurement, comms, and operators. Interview them. Prototype the paper trail.

Don't confuse an offsite with a strategy. Four lines, owners, kill dates, resource shift.

Don't report mural photos. Report funded or killed tests, hours returned, cycle time, one behaviour that moved.

Don't invent a quote from ANZ, Telstra, AGL, DBS, Allens, the NHS, GDS, Services Australia, Capital One, Atlassian, Canva, NAB, Inkeros, or Infinitev. Steal the pattern. Leave their words alone unless you heard them.

If you do the canvas, the brief, and the Monday list at a bank in AU, NZ, HK, or SG, a telco or energy retailer in Australia, a firm like Allens, a public body in AU, NZ, or the UK, a US product org, or a Canadian operator, you will have done more than most labs do in a year. If you cannot, do not book the venue.

Collective Campus exists to make the opposite of theatre true: short, sharp T&E that leaves a residue. A prototype. A decision. A next experiment. Not a mural.